(Invited) Dynamic Evolution for Soft Matter Robot: Integrated Fabrication of Geometric and Motion of Gel-like Robot with Embedded Virtual Link Structure

Abstract

The current 3D printing technology can easily fabricate a body of soft robot composed of various materials. The body of these soft robots has many advantages from the viewpoint of physical interference due to contact using the soft material. Soft robot research in the field of robotics has been reporting a control method to efficiently drive a mechanical structure capable of high freedom movement by combining rigid mechanical parts. However, to transfer these technologies to soft matter science, we are still facing on many problems. As one of the problems, in the academic field of chemistry and robotics, the design concept of soft robot material, geometric structure and movement is not unified. In particular, heterogeneous hybrid materials and link mechanisms are technically separated and discussed. In this study, from the viewpoint of informatics, we integrate the technologies of both by embedding a virtual link structure in the body of soft material. In addition it is an object to create a numerical simulation model for implementing geometry design and physical computation of the soft robot to gradually and simultaneously adapt to external environment. This is because geometry and kinematics for soft robots are not yet established. We have adopted a voxel-based physics calculation library for simulation model. It is open source called Voxelyze. The geometric shape of the soft robot's physical model is created by neural evolution which develops optimization by neural network and evolutionary calculation. When the optimized neural network inputs the coordinate information of the three-dimensional design space, it outputs a three-dimensional physical model composed of voxels of dissimilar materials. Each constitution voxel of the model includes those having an actuator function. It is dominated by the link structure between the corresponding voxels having this actuator function. Our simulation model enables 3D modeling considering the motion of robot which needs time information. In addition, we can find guidelines for simultaneously acquiring structural stability and task performance of soft robot.